Objective
The purpose of this study was to identify peptide classifiers that predict spontaneous preterm birth (SPTB) among women in preterm labor (PTL) and to demonstrate specific protein pathways that are activated in PTL.
Study Design
Serum from 110 women with PTL between 20 weeks and 33 weeks 6 days of gestation was subjected to glycoprotein purification, matrix-assisted laser desorption ionization time-of-flight mass spectrometry peptide profiling, 2-dimensional liquid chromatography tandem mass spectrometry, and pathway analysis. Women were divided into 2 groups: delivery at <34 weeks’ gestation (SPTB group) and delivery at ≥34 weeks’ gestation (PTL group).
Results
Twenty-three peptide masses were identified that discriminated PTL from SPTB in 97% of cases. Fifty-two proteins were present differentially between PTL and SPTB; 48 of 52 proteins were classified into 1 of 4 functional pathways that were involved with PTL: (1) complement/coagulation cascade, (2) inflammation/immune response, (3) fetal-placental development, and (4) extracellular matrix proteins.
Conclusion
Among women in PTL, proteomic analysis of serum peptides and glycoproteins classifies women who will deliver preterm and identifies specific protein pathways at work among individuals with “idiopathic” PTL.
The treatment of preterm labor (PTL) remains one of the greatest clinical challenges confronting obstetricians today; preterm birth (PTB) complicates 12.7% of US deliveries and is responsible for 80% of perinatal deaths. Changes in clinical practice, including the use of progesterone supplementation, nutrient supplementation, and cervical cerclage have all been implemented in specific populations as part of a wider effort to prevent PTB. Despite medical advances, the incidence of PTB continues to increase both nationally and globally.
Clinically, our inabilities to predict PTB and to treat PTL are major obstacles to reducing prematurity. Multiple strategies for the prediction of PTB have been advocated; however, none of these have the predictive accuracy necessary to demonstrate clinical benefits to either the mother or fetus. Likewise, most treatment trials have failed to demonstrate any difference in the duration of pregnancy with the use of tocolytic agents. The origin of these deficiencies lies in our inability to understand the multifactorial nature of PTB and implement a cause-based treatment paradigm into clinical practice. Proteomic profiling, because of its ability to analyze multiple protein signatures concurrently, uses novel methods that ideally are suited for the study of complex conditions, such as PTL.
Current proteomics technologies can probe proteome-wide expression changes between different samples in a high-throughput fashion. Proteins that are differentially present in biologic fluids between groups can then be used to develop sensitive, accurate, and rapid diagnostic assays.
Our objectives in this study were to characterize the maternal serum glycoproteome and maternal serum peptidome in 2 groups of women: those with PTL who delivered at <34 0/7 weeks’ gestation (spontaneous PTB [SPTB] group) and those with PTL who delivered at ≥34 0/7 weeks’ gestation (PTL group). Characterization of the maternal serum glycoproteome was performed to identify key peptide classifiers that would serve as the foundation for a rapid assay to distinguish SPTB from PTL (ie, the prediction of SPTB among women in PTL). Quantification of differential peptide expression and functional categorization of identified proteins was performed to demonstrate specific protein pathways that are activated among women with spontaneous PTL in a cause-specific manner.
Materials and Methods
Subjects
A prospective cohort of 309 women in PTL with suspected intraamniotic infection at the University of Washington from 1991-1997 were identified and underwent cervical-vaginal fluid collection. From this larger cohort, 138 women underwent serum sampling; of these, 28 women had evidence of intraamniotic infection, which left the subset of 110 women with PTL, but no evidence of intraamniotic infection, that was used in this analysis. After informed consent was given, we used a standard collection technique to obtain serum samples from all the women. All women had spontaneous PTL and intact amniotic membranes. Serum samples were processed by a standard protocol, centrifuged, frozen, and stored at –80°C until analysis. The study was approved by the institutional review boards at the University of Washington where the samples were collected and at Oregon Health & Science University where sample analysis was performed.
PTL was defined as the presence of regular uterine contractions accompanied by cervical dilation or effacement between 20 weeks and 33 weeks 6 days’ gestation. Women with PTL who delivered at ≤33 weeks 6 days’ gestation were considered to have had SPTB; women who delivered after 33 weeks 6 days’ gestation were considered as having had PTL with delivery ≥34 weeks’ gestation.
Race/ethnic distinctions were self-described by study participants. Race was assessed to minimize possible confounding between race and protein expression between groups.
Comprehensive proteomic profiles and total serum glycoprotein profiles were derived in 110 women by total glycoprotein purification and digestion and matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) peptide profiling as described later. In a subset of 10 women (n = 5 with SPTB; n = 5 with PTL and delivery ≥34 weeks’ gestation) gestational-age matched serum samples were pooled, depleted of high abundance proteins, and subjected to 2-dimensional liquid chromatography tandem mass spectrometry (2D-LC-MS/MS).
Individual sample analysis (glycoprotein purification, digestion, and MALDI-TOF-MS peptide profiling)
Total glycoprotein purification, digestion, and MALDI-TOF-MS peptide profiling were was performed with 750 μg of total serum protein from each sample with the use of previously described techniques. Discrimination of the resultant serum glycoprotein profiles between SPTB and PTL was performed with ClinProTools (CPT) pattern-recognition software (Bruker Daltronics Inc, Billerica, MA). The recognition capability and cross-validation of the CPT algorithm to discriminate small changes in peptide expression in a complex digest were determined to be 100% through a series of proof-of-principle experiments.
After these preliminary experiments, serum samples were enriched for glycoproteins and subjected to enzymatic digestion. The sensitive (approximately 0.1 μg per spot) and the high-throughput nature of MALDI-TOF-MS was used to obtain peptide profiles on the resulting glycopeptides. Spectral data from quadruplicate analyses for each digested sample were imported into CPT pattern-recognition software, and individual peptide peak statistics were generated at several signal-to-noise levels (2, 3, 4, and 5). The averaged peptide profiles from both subject populations (SPTB and PTL) were then determined. Identification of differentially expressed proteins was performed by separate quantitative time-of-flight analysis.
Pooled sample analysis (2D-LC MS/MS)
Total proteome characterization of individual serum samples with 2D-LC MS/MS is a time-consuming process. Hence, 2 pooled samples, 1 for each condition, were created with a small cohort of serum samples from women with PTL (n = 5) and SPTB (n = 5). Serum was immunodepleted of the 12 most abundant proteins with the use of a Genway Immunoglobin Y-12 column (GenWay Biotech Inc, San Diego, CA). Protein concentration was determined with a detergent compatible protein assay kit (Bio-Rad, Hercules, CA). After protein assay, samples were digested with trypsin, and the digested peptides were separated by strong cation exchange chromatography, as previously published. LC-MS/MS analysis was performed on the cation exchange fractions using a lysine tyrosylquinone (Thermo Fisher Scientific Inc, Waltham, MA) connected to an Agilent 1100 LC (Agilent Technologies, Inc, Santa Clara, CA).
Protein identification
Mass spectra were searched against a composite protein database that contained forward and reversed entries of Swiss-Prot database (version 51.3; Swiss Institute of Bioinformatics, Geneva, Switzerland). Peptide identifications with at least a probability of 0.8 and without any unknown and unexpected modifications are considered as likely to be present in the sample. Protein identifications with at least 2 unique peptide identifications were considered to be present in a sample. Resulting protein identifications and their spectral counts were considered for label-free quantitation.
Peptide quantitation
The total number of tandem mass spectra that were matched to a protein (spectral counting) is considered a label-free, sensitive, and semiquantitative measure for the estimation of its abundance in complex mixtures. The difference between spectral counts of a protein has been used to quantitate its relative expression changes between complex samples. To reduce the false-positive rate, protein entries were further curated before being subjected to label-free quantitation. Curated proteins were then subjected to pair-wise comparisons.
Statistical methods
All proteins and their corresponding spectral counts were subjected to label-free quantitation. Quantitative analysis was performed through pair-wise comparisons between PTL and SPTB samples with either a 2 × 2 χ 2 or Fisher’s exact test. Normalization of spectral counts to account for experimental variability was built into the pair-wise comparisons. The method was automated with an SAS software program (version 9.1; SAS Institute Inc, Cary, NC), and all proteins were independently tested. A protein was considered to be significantly differentially expressed between the samples if the hypothesis has a probability value of ≤ .05 in either the χ 2 or Fisher’s exact test. The fold expression change of differentially expressed proteins was quantified with previously established methods. To reduce false positives from label-free quantitation, only proteins with at least 3 unique peptide identifications in 1 of the samples and showing an expression change of at least ±1.5-fold between the samples were considered to be likely candidates. Analyses of MALDI-TOF-MS glycoproteome spectral data were performed with ClinProTools (version 2.0.319; Bruker Daltonics Inc., Billerica, MA) as we have previously described.
Pathway analysis
The differentially expressed proteins were uploaded into Ingenuity Pathway analysis (2007 Ingenuity Systems, http://www.ingenuity.com ) to categorize the proteins into related functional pathways and networks based on the association between the input proteins and proteins that were present in the Ingenuity Pathway analysis database. The proteins in the networks that were not identified in this study were removed from the networks for better visualization. The networks are displayed graphically as nodes (individual proteins) and edges (the biologic relationships between the nodes).
Results
From the existing cohort, serum samples from 110 women between 20 weeks and 33 weeks 6 days gestation with PTL and intact membranes were identified. Pooled analysis by 2D-LC-MS/MS was performed on a subset of 10 samples (5 SPTB, 5 PTL with delivery at ≥34 weeks of gestation) matched for gestational age (mean, 30.3 ± 1.8 [SD] weeks; range, 28–33 weeks). Comprehensive proteomic profiles and total serum glycoprotein profiles were derived by MALDI-TOF-MS peptide profiling for the remaining 110 women (48 SPTB, 62 PTL with delivery at ≥34 weeks’ gestation). There were no differences between the SPTB and PTL groups in terms of maternal age, race, underlying medical conditions, and pregnancy history. As expected, women in the SPTB group had a significantly earlier gestational age at delivery and shorter enrollment to delivery duration. Of significance, women with SPTB were evaluated slightly earlier in gestation and with more advanced cervical dilation ( Table 1 ). Overall, 24.8% of the women received antenatal steroids, more commonly in the SPTB group (12.9% PTL after 40.4% SPTB; P = .001); and 95.4% of the women received tocolytic therapy (93.4% PTL after 97.9% SPTB; P = .385).
| Variable | Spontaneous preterm birth ≤33 wk 6 d (n = 48) | Preterm labor with delivery at ≥ 34 wk (n = 62) |
|---|---|---|
| Mean maternal age, y | 25.5 | 25.5 |
| Maternal race, n (%) | ||
| White | 30 (62) | 35 (56) |
| African American | 7 (15) | 12 (19) |
| Other | 11 (23) | 15 (24) |
| Plurality of current pregnancy, n (%) | ||
| Singleton | 39 (81) | 60 (97) |
| Twins | 9 (19) | 2 (3) |
| Previous delivery ≤34 wk, n (%) | ||
| No, parous | 15 (31) | 26 (42) |
| Yes, parous | 13 (27) | 11 (18) |
| Nulliparous | 20 (42) | 25 (40) |
| Previous pregnancy complications, n (%) | 15 (31) | 13 (21) |
| Diabetes mellitus at admission, n (%) | 1 (2) | 4 (6) |
| Uterine anomaly, n (%) | 2 (4) | 3 (5) |
| Hypertension at admission, n (%) | 3 (6) | 0 |
| Mean gestational age, wk a | ||
| At enrollment | 29.4 | 31.2 |
| At delivery | 30.6 | 37 |
| Mean days between enrollment and delivery | 10.6 | 40.4 |
| Mean cervical dilation at enrollment, cm a | 4.1 | 3.3 |
a P < .001, comparison of groups by nonparametric analysis of variance (continuous variables) or χ 2 test (categoric variables).
With the use of the total glycoproteome digests, the average discrimination power over 16 computed models was 97.3% for PTL and 97.0% for SPTB. Twenty-three peptide masses that are used commonly by the CPT classifier to differentiate between SPTB and PTL are shown in Table 2 . Fifty-seven percent of the CPT peptide classifiers were mapped to 8 proteins. The false discovery rate of this technique was estimated at 3%. The averaged MALDI-TOF-MS glycopeptide profiles from SPTB and PTL samples are shown in Figure 1 , A. Magnified examples of 2 differentially abundant peptides at 980 dalton (haptoglobin) and 1530 dalton (serotransferrin) are shown in Figure 1 , B.
| Peptide mass | P value b | Peptide sequence | Swiss-Prot accession | Protein description |
|---|---|---|---|---|
| 980.48 | < .0001 | VGYVSGWGR | P00738 | Haptoglobin |
| 1529.7 | < .0001 | KPVEEYANCHLAR | P02787 | Serotransferrin |
| 1833.87 | < .0001 | VFSNGADLSGVTEEAPLK | P01009 | α-1-Antitrypsin |
| 1650.77 | .0001 | |||
| 1195.53 | .0007 | DSGFQMNQLR | P02787 | Serotransferrin |
| 1129.62 | .0012 | RLWWLDLK | P02790 | Hemopexin |
| 1283.54 | .0020 | EGYYGYTGAFR | P02787 | Serotransferrin |
| 1723.92 | .0023 | QKVEPLRAELQEGAR | P02647 | Apolipoprotein A1 |
| 1213.59 | .0023 | |||
| 1878.86 | .0042 | EYVLPSFEVIVEPTEK | P01024 | Complement C3 |
| 1741.78 | .0052 | |||
| 3282.38 | .0140 | |||
| 1647.75 | .0141 | |||
| 1478.71 | .0179 | MYLGYEYVTAIR | P02787 | Serotransferrin |
| 2833.46 | .0203 | |||
| 1855.94 | .0307 | FNKPFVFLMIEQNTK | P01009 | α-1-Antitrypsin |
| 1895.88 | .0330 | |||
| 1743.75 | .0360 | |||
| 1755.85 | .0504 | SDPVTLNVLYGPDLPR | P11464 | Pregnancy-specific β-1 glycoprotein 1 |
| 2239.98 | .0516 | |||
| 2493.2 | .0739 | |||
| 1323.63 | .0872 | DSAHGFLKVPPR | P02787 | Serotransferrin |
| 1752.96 | .1300 | YVGGQEHFAHLLILR | P02763 | α-1-Acid glycoprotein 1 |
a Peptide fingerprints determined by matrix-assisted laser desorption ionization time-of-flight mass spectrometry;
b P value corresponds to differential abundance between spontaneous preterm birth and preterm labor.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
